Stephen Harmston

The world is being flooded with AI-generated content. Much of it is great and very useful but some is used for misinformation, fraud and social engineering. Current approaches to detection tend to rely on single models which are brittle when faced with new types of content. We have built a new type of approach which is multi-model and multi-modal.

I've tried quite a lot of translation systems and they all seem to struggle with slang - how do you deal with that - have you tried translating slang into one language and back - was it effective?

Can you set bespoke 'conversion' goals (for example exploring other areas of a page or moving to a particular page from the home page) as well as standard conversions such as sign ups?

Good idea. Interested to see how it develops and how scales.